[00:38:29] * bd808 off [08:45:04] morning [12:55:23] Good morning! [13:09:49] o/ [14:57:26] very easy review related to what we just talked: https://gerrit.wikimedia.org/r/c/operations/puppet/+/979056/ [15:00:35] nice [15:09:23] added it to the task :) https://phabricator.wikimedia.org/T328502 [15:09:48] I think icinga did not notice yet though xd [15:12:21] i'm already tracking progress in subtasks like https://phabricator.wikimedia.org/T345294 :) [15:14:19] awesome, I'll let you do then [16:08:15] I accidentally pushed https://gitlab.wikimedia.org/repos/cloud/toolforge/alerts/-/commit/decc112f10d04813710a67e32f5cbb6669ae5932 without review :/ [16:08:34] anyone want to have a quick look or should I revert + re-submit as a MR? [16:12:59] it's small enough that I'm fine with not reverting, I'm doing a quick review now [16:20:28] looks good, my only comment is that the wording is a bit similar in the two alerts ("close to" and "almost"). when the thresholds are lowered in the future, I would also reword the messages [17:24:50] Here, at long last, is the diff in drive errors: https://phabricator.wikimedia.org/T348643 [17:25:07] summary: small increases on some but not all tested drives [17:27:18] dcaro: thanks for editing the buildpack wiki about multi-stack builds using "pack". I added an extra note under https://wikitech.wikimedia.org/wiki/Help:Toolforge/Build_Service#Using_Node.js_in_addition_to_another_language [17:27:46] 👍 [17:28:04] I'm writing a reply and testing stuff locally [17:31:26] I've just added a reply, sorry I didn't see your message here :) [17:31:36] Feel free to add more info in the thread [17:47:36] aaaaaaaah. we have a tools-k8s-worker-100 now, and `kubectl sudo get node` is sorting that before tools-k8s-worker-30 instead of after tools-k8s-worker-99 like I would want it to [17:56:33] `| sort -V` might help [17:58:10] This reminds me of a previous discussion about 0-padding node names... [17:58:15] unpadded numeric sort is evil [17:58:16] which went nowhere :) [17:58:43] (it was this https://gerrit.wikimedia.org/r/c/operations/cookbooks/+/868803 ) [17:59:37] ihmo for things that we have a high number of (like k8s nodes) we should just give a few random alphanumeric characters and call that good [18:03:52] I'd say for anything we don't care the sequence creation of (most stuff probably) [18:22:58] It's moderately useful to have sequential IDs because we can notice right away that a set were built together. [18:23:06] Although of course the launch time also works for that [18:27:52] kinda, might depend, as you can build 17 today, and 18 in a month [18:28:36] * dcaro off [18:56:35] * bd808 lunch [19:09:06] taavi, dhinus, like this wording OK? https://gerrit.wikimedia.org/r/c/openstack/horizon/horizon/+/982454/1 [19:28:23] sgtm! [22:57:19] well I had a nice 6-month run when horizon-on-docker behaved the same on my laptop and in production :(